M.S. Applied Data Science - Capstone Chronicles 2025

12

The persistence of high prevalence even among individuals with good lifestyle scores in the multiple-medication categories challenges the initial hypothesis that lifestyle-only models could match the performance of medication-inclusive models. Instead, these results support the view—also seen in earlier plots—that incorporating medication data adds meaningful predictive value. Finally, Figure 6 compares mean scores for lifestyle effort, physical activity, and dieting likelihood across the four medication classes. Those not taking medications scored slightly higher on lifestyle effort (just above 1.75) and activity level (around 1.5) than medicated groups, but all scores were relatively low overall. Dieting likelihood was uniformly low (< 0.25), with the highest values among those taking both

MetSyn-related and unrelated medications, potentially reflecting a reactive change in diet after multiple diagnoses. These patterns indicate that while healthier lifestyle behaviors are somewhat more common in the no-medication group, the differences are not large enough to sharply distinguish metabolic syndrome risk when clinical factors are also considered. Taken together, these exploratory findings suggest that while medication use is strongly correlated with metabolic syndrome status, lifestyle effort and income level show weaker associations. Dietary patterns, physical activity, and self-reported health behaviors, though potentially valuable in combination with clinical measures, appear less decisive on their own.

Figure 6 Comparison of lifestyle effort, activity levels, and dieting behavior across medication classes. The "None" group had the highest mean scores, although overall levels were low.

163